How LYTT’s machine learning algorithms are changing the oil & gas industry

James Ramsay, Data Scientist @ LYTT, May 11 2020

In this article, I’d like to talk about one of the major energy industry challenges that LYTT has solved in the last year as well as describing my experience as a Data Scientist since joining LYTT.

LYTT is a rapidly growing tech start-up delivering ground-breaking analytics and technologies to tackle complex energy challenges. Started in 2019, we have already generated several $100 million of value for our clients and we are excited to see what we can achieve in 2020.

All that aside, the most important thing to take away from this post is how to pronounce LYTT correctly… It’s “light” not “lit”, OK?!?

My journey began with studying Maths at Oxford University followed by an MSc in Petroleum Engineering at Imperial College. I then worked for a number of years in oil and gas consulting where I became gradually more involved in Data Science as the field grew. When a Data Science job opportunity presented itself at LYTT — exciting new start-up with a role that perfectly matched my skill set — I was eager to apply and very happy to be accepted.


The challenge
One of the main challenges that LYTT aims to address is understanding the subsurface fluid flow in oil and gas production wells. Oil wells typically produce a mixture of oil and water, where is water is an unwanted byproduct that is expensive to clean and dispose of.

However, it difficult to detect where in the subsurface the produced water is coming from. Without this knowledge, it isn’t possible to intervene and stop this water production, which is necessary step for maximising oil production.

challengeWhat depth is the water production coming from?


The solution
LYTT’s innovative solution involves deploying fibre optic cables — just like the ones we use to send data for broadband — inside oil wells. By pumping light up and down the fibre and measuring changes in the light’s character, it is possible to measure both the sound and temperature at each point along a the fibre.

LYTT’s bespoke machine learning models are able to use the sound information collected by fibres deployed inside oil wells to detect the subtle differences in the acoustic signatures of oil flow and water flow.

SolutionLYTT’s analytics detect at which depths water production is coming from.

These insights allow our clients to take remediation measures to prevent water production and consequently maximise their oil production!

Building models that can reliably distinguish between oil and water has been extremely challenging. We have relied on advanced signal processing and modern machine learning techniques coupled with high-speed computing to enable this. To complicate matters further, no two oil wells sound quite the same. Designing a set of acoustic features that generalise from one well to the next has been one of our most significant achievements.

LYTT’s offering
LYTT’s algorithms, such as the water detection application described above, have been deployed over 100 times in oil and gas fields across the world.

LYTT’s is able stream our fibre-derived insights from wells in remote to locations to a cloud-based visualisation platform, all in real-time. This is no mean feat; a single well with a fibre optic installation generates 300 MB/s of data! But the ins-and-outs of handling these volumes of data can be the subject of another blog post.

I hope you’ve enjoyed reading.

This website uses cookies. Click here to learn more. That's Fine